JPH10162020A - How to browse the image database - Google Patents

Abstract

(57) [Summary] [Problem] To present an image with a small number of presentations by narrowing down images required by a user step by step. SOLUTION: Based on a feature extracted from an input image, a category generation unit 103 divides the image into categories. The link generation unit 104 generates a link between images having similarity even in images between different categories. Image presentation means 10
5 presents a representative image of each category to the user.
For the image selected by the user, the image of the sub-category and the linked image are further presented.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for browsing an image database for displaying images desired by a user while narrowing down the images in a stepwise manner.

[0002]

2. Description of the Related Art As a conventional technique for retrieving a face image, a landscape image, and the like, there is a similar face image retrieval apparatus described in Japanese Patent Application Laid-Open No. 7-302337, for example. In this search device, face images are made into parts and registered in a database,
The input face image is compared with the face images registered in the database, and persons are selected in similar order and presented to the user.

[0003]

However, according to the above-mentioned conventional method, at the time of database registration, the user needs to have a clear image of each face part feature in advance, and once selected and presented. If you are dissatisfied with the results, you need to start over.

The present invention has been made in view of the above circumstances, and an object of the present invention is to provide an image database browsing method for presenting images with a small number of presentations by narrowing down images required by a user step by step. Is to provide.

[0005]

According to the first aspect of the present invention, there is provided a method of browsing a predetermined image stored in an image database, the method comprising:
All images are divided into a plurality of categories according to one or more features, and for the divided different categories of images, the image database is created by generating links between similar images, and a sample of each category is created. Displaying an image, and displaying again a sample image of a subcategory in a category to which the image designated and selected from the displayed images belongs, and a similar image linked to the selected designated image. I have.

In the invention described in claim 2, the link to be generated is such that when a predetermined feature is similar in the one or a plurality of features, a link is generated between the images regardless of other features. Features.

According to the third aspect of the present invention, when there are a plurality of images designated to be selected, a plurality of categories to which the selected images belong are regarded as pseudo one category and a new sub-category is generated. It is characterized by.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be specifically described below with reference to the drawings. FIG. 1 shows the configuration of an embodiment of the present invention. In the figure, 101 is an image input means, 10
2 is a feature extracting means for extracting various features from the input image, 103 is a category generating means for categorizing an image group, and 104 is a link generating means for generating a link using some of the extracted features. 105, an image presenting unit such as a display; 106, an instruction input unit from a user such as a mouse; 107, a data storage unit for storing input images and various information during processing; 108, an overall control unit; Is a data communication path.

FIG. 2 is a processing flowchart of the present invention. Hereinafter, the operation of the present invention will be described according to the processing shown in FIG. First, a document image is obtained by the image input means 101 (step 201). This image input means is, for example, a scanner, a facsimile or the like, and may obtain an image from another device via a network.

Next, the features of the input image are extracted by the feature extracting means 102 (step 202). here,
As a feature to be extracted, for example, in the case of a face image, a shape feature such as an outline and an eye position, a color feature, and the like are extracted. Next, the category generation unit 103 categorizes the image group (Step 203). This is divided into categories by performing clustering in the feature space as shown in FIG.

Referring to FIG. 3 as an example, in an interrogation having three characteristic axes X, Y, and Z, clusters are obtained such that intra-cluster variance is small and inter-cluster variance is large. In FIG. 3, four clusters 301 to 304 are obtained. Each cluster in this feature space corresponds to each category of the image group. Then, by performing clustering again in each cluster, 305, 3
A sub-cluster (category) as shown at 06 is generated.

Subsequently, the link generation means 104 generates a link (step 204). This will be described with reference to FIG.
By performing clustering in the feature space as described above, there are cases where two images belonging to different clusters have a very close relationship (that is, are similar) except for a certain feature. Clusters 401 and 402 are examples, and Z
The relationship is very close if the characteristics of the axis are ignored. Therefore, instead of using all the features, a link is generated between images that are very close to each other in some features (this is determined by a threshold or the like).

An image database is created and prepared by the above processing. Hereinafter, processing when the above-described image database is used will be described. First, the image presentation means 10
5 selects one representative image of each category obtained in step 203, and presents (or displays) it to the user as shown in FIG. 5A (step 205). Here, as a representative image of each category, an image closest to the center of gravity in the cluster may be selected. Then, the user selects one image closest to the desired image using the instruction input means 106 (step 206). For example, FIG.
In (a), it is assumed that the image 505 has been selected.

An image representing the sub-category of the category to which the selected image belongs and an image linked to the selected image are searched (step 207), and presented again as shown in FIG. 5B (step 205). ). In the example of FIG. 5B, the image 505 is displayed on the image 511, and the image 505 is displayed.
4 sub-categories in the same category as and image 50
Assuming that there are two images linked to 5, the representative images of the sub-category are presented as 512 to 515, and the linked images are presented as 516 and 517. The above processing is repeated until the user instructs termination (step 206).

In FIG. 5A, the user may select a plurality of presented images. In such a case, a method may be adopted in which the selected plurality of categories are regarded as one category in a pseudo manner, and a subcategory is newly generated in the category.

Note that the present invention is not limited to the above, and can be realized by software. When the present invention is realized by software, CP
A general-purpose processing device including a U, a ROM, a RAM, a display device, a hard disk, a keyboard, a CD-ROM drive, etc. is prepared, and a program for realizing the processing functions of the method of the present invention is stored in a recording medium such as a CD-ROM. Has been recorded.

[0017]

As described above, according to the first aspect of the present invention, it is possible to narrow down the images desired by the user stepwise by dividing the categories stepwise.
Also, by using the link, an image exceeding the category is also presented, so that even if the selection of a higher-order category fails, a desired image can be selected without starting over from the beginning of the image presentation.

According to the second aspect of the present invention, it is possible to present an image having similarity to a user even if the image exceeds the category.

According to the third aspect of the present invention, it is possible to select a desired image by selecting a plurality of images even if it is difficult for the user to select one of the presented sample images. Becomes

[Brief description of the drawings]

FIG. 1 shows a configuration of an embodiment of the present invention.

FIG. 2 shows a processing flowchart of the present invention.

FIG. 3 shows an example in which an image group is classified into categories in a feature space.

FIG. 4 shows an example of generating a link between images.

FIG. 5A shows an example of an image presented to a user, and FIG. 5B shows an example of an image presented again.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 101 Image input means 102 Feature extraction means 103 Category generation means 104 Link generation means 105 Image presentation means 106 Instruction input means 107 Data storage unit 108 Control unit 109 Data communication path

 ──────────────────────────────────────────────────の Continued on the front page (72) Inventor Tsukasa Kochi 1-3-6 Nakamagome, Ota-ku, Tokyo Inside Ricoh Co., Ltd.

Claims (3)

[Claims] 1. A method for browsing a predetermined image stored in an image database, wherein all images are divided into a plurality of categories according to one or a plurality of characteristics, and similar images are classified into different categories. Creating the image database by generating links between the images to be displayed, displaying sample images of each of the categories,
Browsing of an image database, wherein a sample image of a subcategory in a category to which an image selected and instructed from the displayed images belongs and a similar image linked to the selected image are displayed again. Method. 2. The link to be generated according to claim 1, wherein a link between the images is generated when a predetermined feature of the one or more features is similar, irrespective of other features. Browsing method of the described image database. 3. The method according to claim 1, wherein when there are a plurality of the images designated to be selected, a plurality of categories to which the selected images belong are regarded as one category, and a new sub-category is generated. The browsing method of the image database according to 1.